Tuesday, April 10, 2012

Optimizing the transfer ratio

At the previous post I made the following comment:
This means that the transfer ratio is reciprocal to the complexity of the pattern. In order to maximize transfer we must minimize complexity. I doubt that.

The reason that I have doubts is that I don't have proof that it actually works.

It hasn't always been that way. Especially my experience with Troyis once seemed to indicate that it might work. The idea is very elegant and tempting, of course. Learn to do the skills and pattern recognition you need almost every move and do it as fast as possible.

With this idea in mind I have spent 1000 hours or more with the following exercises:
  • Low level drills of DLM.
  • Exercices of CD Maurice Ashley teaches chess.
  • 3 board vision exercises by Fritz.
  • Knightmoves by playing Troyis.
  • Tons of low level tactics.
Tragically these exercises didn't have a measurable effect. The main problem is the relevance of the skill in relation to the outcome of the game. Or to the solution of a complex problem, for that matter. It is nice if you can move around a knight in your head at lightning speed, but it very rarely decides a game.

Aox is a firm believer of this method. I wish him all the luck. In order to make it work you have to discover a new skill. Or new category of patterns, for all you pattern afficionados out there. Patterns with a high transfer ratio and a high relevance for the outcome of the game/solution of the problem.

I have invented such new category of patterns: the invisible topology of the board. Is it frequent: you need it every position. Is it relevant to the outcome: you will find no tactic where it plays no role.

To be perfectly honest with you: I lack the energy to exercise it without any proof that it actually works. I hope you can understand that.


  1. What a nice door stopper, O my!


  2. Those can be remarkably useful. That is often underestimated:)

  3. Like you, tempo, I dont believe that training the veeery easy stuff gets you anywhere. However, I also believe that my 3 boardvision training (attack, defence and check training) was actually beneficial. To some extend that is. I reached 47 or so attacks per minute, I think that is good enough, and a lot better than the value I started with (which was 8 or 12).

    The main problem I have with this method is, that it is not adding any intelligence. (But that is only half true. I certainly added some intelligence by pushing my original board vision from 8 attacks to 47 attacks per minute).
    Well, it is a matter of diminishing returns.

    The same will be true later for the range I currently train (CT Blitz rated puzzles between 1300-1475).
    I let my feeling guide me, and I decided to drop the range 1150-1300. Before my training range was 1150-1475, but since January or so it is only 1300-1475.
    This does not mean, that my time with the range 1150-1300 was wasted. But for me - I "sucked" the intelligence out of this range, and felt it was time to move on.

    I notice patterns I trained in first-timer-puzzles, which gives me the feedback, that I my training has an impact on my chess thinking.

    For aox, I am not sure if you misinterpret him here. Aox believes that gaining speed is beneficial. Me, too, and certainly you, too.
    The question remains how to speed up? Through repetition of what we learned (automation of the added intelligence).
    That is hopefully the key to success.
    If 3 board vision adds still considerable intelligence, then it is still a good training. If it adds intelligence - this answer can only be found within each of us. You need to try. If you "feel" you learned enough, then it is time to move on. The 100% will never be achieved. But after we learned 90% of what could be learned, I believe it isnt worth to continue it. Like I stopped the 3 board visions and the 1150-1300 range for that very reason - hardly intelligence anymore added, in comparison with a different range.

  4. I do not believe “the simpler the better here.” However, it makes sense to learn as many lessons as you can from simple problems. You also have to be realistic about the time you have to look at tactics per move, when the position does not scream tactics. Half a minute? A minute? You also have to be realistic about what you would spot in a real game. You have lots of clues when solving a problem that you do not have in a game. Another practical point is that you can pretty much guarantee to get quicker at the easy stuff, until diminishing returns set in. Progress is less certain with the harder stuff.

  5. I did an experiment with a bigger set of low rated mate - problems and did solve them in random order.( i already solved several 10000, maybe 150000? tactic problems before). It was not possible for me to improve at this set. I did some analysis why this happend. If i dont see a problem again within 8 days then i could see no improvement in speed when i sloved it a second time. I did forget!
    The set was to large to see the same problem early enough to gain speed and the problems where not related enough to benefit enough from transfer effects.
    So repetition of some type has to be early! If you have to many pattern and not enough repetition you cant improve.
    Just doing " Tons of low level tactics" does realy have no effect. Without repetition at the right moment the effect of having done a problem is zero.

    My new experiment is very promising. Now repetition is garantied (CT-SRS), (but i think still not "best"), transfer seems to happen.

    Bright Knights experiments with low rated problems and spaced repetition are proving transfer and improvement too.

    Just because something which cant work dont work, dont mean that something which does work, dont work.

  6. @Aox,
    Just because something which cant work dont work, dont mean that something which does work, dont work.

    Feel free to underestimate my efforts, my friend. I haven't always worked with big sets. SRS might be best, but there are other methods for memorizing that works. I happen to have a very good memory. In my haydays I had memorized over 20,000 positions. I could do Ftacnik's 1000xcheckmate in under 2 hours (7 secs per problem) and could solve several other sets of 1000-1500 problems at that speed.

  7. After the initial boost - which I had when I didn't repeat at all - it stalled. I continued it for 3 extra years while stalling to be sure that the law of diminished returns had really set in.

    You see, when you start with tactics and work hard you will improve 250 to 350 points anyhow. No matter which regimen you follow. That's why I don't consider the proof of you and Aox a proof just yet. Only when you stall and manage to launch again we are talking about proof. Or when you improve more than 350 points while your starting point is above 1500.

  8. Did you stop improving OTB or did you stop improving tactics? I think only doing tactics cant be enough to improve in OTB "for ever".
    A checkmate-training has a low effect at OTB (says Dan Heisman, i am not shure about that) but it would have a big effect at chesstempo, 40% of the problems are checkmate problems.

    If Kmelnitsky is right then an improvement in tactics of 100 points will give you an OTB improvement of 40 points. But i am shure that is only true if all chess skills are more ore less equal developed. If you are already a tactical "genius" then more tactic dont help anymore.
    Then you should start to use the methods you developed for tactic trainng at Strategy, endgame, opening and so on.

    I had a plateau for almost one year ( after maybe 150000 attempts ) and i did improve by using repetition. I think plateauing is only possible if the quantity*quality of what you get is of the same size as the quantity*quality you lose.
    I think SR can help to keep memory and i think memory and acess to memory is the main key.
    But it is possible that at a certain level you only aquire skills by loosing skills . There are some papers about that phenomenon.
    If this method dont work, and i will see that the next ? 12 ? months, then i will give up tryining to improve and look for a new hobby. Mabye alcohol...

  9. Did you stop improving OTB or did you stop improving tactics?


    I don't have the impression that I had reached my tactical maximum then. Not by far.

    I believe the devil is in the details. The details of the method, that is. That's why I continue. Of course we'll find it. It is a matter of elimination and creative thinking. And time and effort, of course.

  10. If i dont see a problem again within 8 days then i could see no improvement in speed when i sloved it a second time.

    If you repeat that problem every 8 days, I am sure that you would improve. Early repetitions make the later repetitions easier, but do not contribute much to long term memory.

  11. @Bright Knight
    I did plateau at this set. The over all improvement was 0. What i did improve at "one side" i did lost at the other. I did gain some speed at problems i did see less than 8 days ago and i lost speed at the other problems. ( 2,3,.. repetition to late ). This experiment did take me a while but i was learning from it.

    ( I am shure the number of days (8) is a function of the rating of the problems and other factors)

    Its like filling a pool(brain) with holes all over(forgetting) with water(information). Its hard to fill it up. If you just put water in it, with constant speed, you will reach a certain level and never get higher anymore.
    Its necessary to close the holes! which means you need to work on the forgetting speed.
    That can be done by repetition ( as you already know )
    See http://en.wikipedia.org/wiki/Forgetting_curve


    Hmm, thinking about that ,..
    Some easy math may show if this strategy does work ( and how much )

  12. As I have said before, if Tempo repeated what he did before, he should repeat his former success. However, unless the performances quoted above were specific to those problem sets and short lived, he appears to have been much stronger at simple tactics than others of his overall strength. What was letting him down? Dan Heisman's "thought processes" spring to mind.

  13. @BK,
    well, yes, but that is just a matter of maintenance. I have not the slightest doubt that I will get those 150 points back soon when I set myself to that.

    But then I will be at the same plateau again. That's what bothers me and keeps me busy with the things I do. I look to the future. It is more logical to me to put effort in the quest for a better method than to continue in a way I know is suboptimal by far.

  14. I doubt whether the difference between the optimal training method and a good one amounts to very much. Provided that you avoid serious wastage of time, or neglect something important, I expect that what matters is how much work you put in.

  15. @BK,
    Provided that you avoid serious wastage of time, or neglect something important, I expect that what matters is how much work you put in.

    That's my point. We neglect a few important things.

  16. I meant neglecting big things, like tactics, or the endgame. Small things probably iron themselves out.